Re: Statistical modeling of novel chemical systems( AKA Chemist with lots of stats questions)


Old Mac User wrote:
If I'm reading this correctly, you have one controllable variable
"X" and one dependent variable (also called a 'response') "y" and
you are trying to establish a relationship between "y" and "x"
by setting "x" to different values and finding the corresponding
value of "y" by experimentation.

If this is right, then it would be helpful to have the data.
Without some data I would be in the position of writing long
posts to discuss the possibilities. Also, there may be more
to this than I've understood.

If you want to do so, feel free to contact me at the following
e-mail address: hedging77 followed by @yahoo.com
If I write this in the usual way, Google Groups will fix it so
it can't be copied by spammers.

Just fyi, I'm a chemical engineer and also a statistician.
44 years of experience in the chemical industry, now "retired"
(sort of).

Don't let that fellow get you down!! OMU


Sorry to take so long to get back to you, I was at a wedding over the
holiday. I think I have managed to figure out the problem. It was too
simple, all I did was figure out the mean and stdev of my data at each
time and then fit the curve by minimizing the ratio of the stdev to the
distance from the curve. That way the system corrects for the
exponential growth rate and the error at each time. And its easy
enought to calcuate the confidence interval at each time given that
informantion.

I heard back from the corporate stats guy and will be meeting him this
afternoon, because apparently he has a method of calculating true
confidence intervals on a continous curve, but the software isnt excel
friendly. It would be great to get the parameters of the equation
rather than just the percent match and the parameters at each point in
time.

I will email the data this afternoon hopefully, after I speak to the
stats guru, I am hoping that you can explain what exactly he is doing
to the data.

Ghostwriter

.